Abstract

Results and Discussion

We used gastric mill (GM) neurons of the STG of the crab, Cancer borealis. GM neurons were isolated from all known neuromodulatory and synaptic inputs (24) and were silent under control conditions. The dynamic clamp was used to establish reciprocal inhibitory synaptic conductances (g_syn) between 2 GM neurons and to add an artificial hyperpolarization-activated inward conductance (g_h) to each neuron (Fig. 1A). Depending on the strength of the inhibitory synapses, and the amount of g_h, these 2-celled circuits displayed a variety of circuit outputs (Fig. 1 B and C). In each experiment, g_syn and g_h were varied over a large range (typically from 10 to 100 or 115 nS) in steps, thus generating a map containing 49–64 different combinations of synaptic and intrinsic parameters (Fig. 1B). The map in Fig. 1B is color-coded to illustrate the circuit behaviors: green indicates circuits in which 1 neuron was active and the other silent, blue indicates circuits in which both neurons were active but firing asynchronously, red indicates map points showing stable half-center oscillator behavior (24, 25) (alternating bursts of activity), and black shows map points at which both neurons were silent. Fig. 1C shows recordings from the circuits at the indicated map points.

Open in a separate window

Fig. 1.

Circuit output depends on both synaptic and intrinsic conductances. (A) Schematic shows circuit constructed from 2 GM neurons with the dynamic clamp. The dynamic clamp creates a symmetrical reciprocally inhibitory synaptic conductance (g_syn) and an artificial intrinsic membrane conductance (g_h). These maximal conductances were varied between 10 and 115 nS. (B) Map of g_syn–g_h parameter space for coupled GM neuronal circuits. Black, green, blue, and red describe the circuit behavior as silent, asymmetrical, spiking, and bursting, respectively. (C) Intracellular recordings show voltage traces of representative circuit activity patterns. Colors are as in B; boxed points in B correspond to the parameter values of the traces in C. (Vertical scale bar: 20 mV; horizontal scale bar: 5 s.)

Both serotonin and oxotremorine act directly on isolated GM neurons. Serotonin is known to modulate a number of different ionic currents in STG neurons, including I_h, Ca²⁺ currents, and K⁺ currents (26–30). When applied to isolated GM neurons (n = 46), 10⁻⁶ M serotonin had a small, but statistically significant, effect on resting input impedance (control 22.3 ± 8.2 MΩ versus 25.3 ± 9.6 MΩ in serotonin, P < 0.003) and a small, but statistically significant, change in resting potential (control −62.3 ± 5.8 mV versus −59.9 ± 5.0 mV in serotonin, P < 0.001). The most dramatic effect of serotonin on the intrinsic properties of the GM neurons was a hyperpolarizing shift in the spike threshold in response to depolarizing ramps of current (control −42.4 mV ± 5.3 mV versus −51.0 mV ± 8.1 mV, P < 0.001). The consequences of these changes in intrinsic properties are seen in the response of GM neurons to injected current. Fig. 2A shows the response to the same depolarizing current pulse in control saline and serotonin. Fig. 2B shows a full FI (frequency/current) curve for the same neuron shown in Fig. 2A. Fig. 2C shows an average FI curve taken from fits to the FI curves of 22 neurons in control and serotonin. Note that the 2 curves are parallel, with the threshold for firing lower in serotonin.

Open in a separate window

Fig. 2.

Serotonin excites GM neurons. (A) Voltage recordings from a GM neuron, which was given the same current pulse (black) in control (blue) and serotonin (red). Horizontal lines: − 40 mV; 0 nA. (B) Frequency/current (FI) plot from neuron in A. (C) Mean FI curves calculated by fitting a line to each individual cell's (n = 22) FI curve, and then obtaining a mean slope and a mean y-intercept to create an “average” FI curve.

After the construction of a control map, 10⁻⁶ M serotonin was applied to each preparation, and then a second map was made, allowing the examination of the effects of serotonin on 49–64 different circuits within each experiment. Fig. 3 shows the effects of serotonin on a typical preparation. In Fig. 3A, the maps are shown in 3 dimensions with the frequency of the half-center oscillatory rhythms (red dots) on the z axis. In control saline, the half-center frequencies were low. In the presence of serotonin, the half-center region expanded, and the frequency of many of the half-centers increased. Alternating single spikes were considered half-center activity [consistent with the tradition in experimental studies of locomotion (31, 32)], and many of the high-frequency points in Fig. 3A and the remaining figures consist of alternating single spikes.

Open in a separate window

Fig. 3.

Serotonin expands the parameter sets that produce alternating bursts of activity and increases network frequency. (A) (Left) Control saline. 3D plot showing half-center frequency as a function of g_syn and g_h in circuits with half-center activity. (Right) In 10⁻⁶ M serotonin. (B) Sample traces in control and serotonin. (Upper) g_syn = 55 nS and g_h = 40 nS. (Lower) g_syn = 70 nS and g_h = 115 nS. Horizontal line on traces indicates −50 mV. (Vertical scale bar: 20 mV; horizontal scale bar: 10 s.) (C) (Left) Distribution of bursting circuits in control. Grayscale shows the fraction of half-centers at each g_syn–g_h parameter set (n = 17). (Right) Serotonin (10⁻⁶ M) increased the fraction of bursting circuits at many g_syn–g_h map locations and expanded the range of parameters that produced bursting activity within the sampled parameter space. (D) Serotonin (10⁻⁶ M) increased the percentage of the map that contain bursting half-center circuits in each of 17 experiments (paired t test: P < 0.001). (E) Serotonin (10⁻⁶ M) increased the average frequency of the bursting half-center oscillator circuits in each experiment (paired t test: P < 0.001).

Fig. 3B shows recordings from 2 of the map points, at the parameters indicated in the figure caption. The top example (Fig. 3B) shows a circuit that was silent in control saline and became an oscillator in serotonin. The bottom traces of Fig. 3B show a circuit that was bursting in control saline and increased its burst frequency in serotonin. In general, the spike frequency and number of spikes in the burst also changed in serotonin. Fig. 3 C–E summarizes data pooled from 17 experiments. Fig. 3C shows that some parameter regions were more likely to produce half-center activity across all of the preparations. Fig. 3D demonstrates that more circuits produced stable alternating half-center activity in the presence of serotonin than in control saline. Fig. 3E shows the mean change in burst frequency of all of the rhythmic half-centers in all 17 experiments. These data demonstrate that, on average, serotonin significantly increased the burst frequency of the half-center oscillators (P < 0.001). The amount of this increase varied considerably across preparations, although the population effect was robust.

Despite the fact that the effects of serotonin on the pooled data are quite consistent, an examination of all of the individual half-center oscillator circuits shows that not all half-centers increased their frequency in serotonin. A few circuits responded anomalously, with a decrease in burst frequency (Fig. 4A). Fig. 4B shows the frequency in serotonin as a function of control frequency for 868 networks taken from 17 preparations. Most of the points are above the dashed identity line, but some are found below the identity line. There was a large range in the burst frequency increase seen in serotonin, and some parameter combinations showed large frequency increases. In Fig. 4B, asterisks are used to show those networks with alternating spikes and circles indicate alternating bursts. We then asked whether the anomalous behavior in serotonin was at similar parameter map positions across experiments. In 5 experiments, no anomalous points were seen, in 5 experiments 1 map point responded anomalously, and in 7 experiments 2 or more map points responded anomalously. Fig. 4C shows that there were no preferential positions (absolute or relative) of the map points that responded anomalously to serotonin. However, alternating single spike bursters showed characteristic map locations.

Open in a separate window

Fig. 4.

Mean serotonin-mediated increase in burst frequency masks serotonin's frequency reduction of a small number of circuits. (A) Example traces of a circuit that slowed in serotonin (g_syn = 85 nS and g_h = 85 nS). Horizontal line on traces indicates −50 mV. (Vertical scale bar: 20 mV; horizontal scale bar: 20 mV). (B) Circuit frequency in 10⁻⁶ M serotonin as a function of control frequency. Asterisks indicate networks with single-spike bursts in either control or serotonin conditions. Dashed line indicates x = y. (C) Location of g_syn–g_h parameter values of circuits that slowed (downward arrows) in 10⁻⁶ M serotonin in 3 experiments. Asterisks indicate networks with single spike bursters as seen in B.

To determine whether a different neuromodulator would produce effects similar to those of serotonin, we repeated these experiments with oxotremorine (Fig. 5). In STG neurons, oxotremorine activates a distinct current from the ones activated by serotonin. Specifically, oxotremorine activates the same modulator-activated, voltage-dependent inward current (I_MI, formerly called the “proctolin current” or I_proct) as do many neuropeptides (33–35). Because of its voltage dependence, this current is largely off when the neuron is hyperpolarized, but it increases excitability as the neuron depolarizes. When applied to isolated GM neurons (n = 19), oxotremorine had no effect on resting input impedance (control, 28.0 ± 12.3 MΩ versus 27.1 ± 7.6 M Ω in oxotremorine, P = 0.73). Like serotonin, oxotremorine produced a small, but significant, depolarization of the resting potential (control, − 59.5 ± 4.7 mV versus −56.0 ± 4.3 mV in oxotremorine, P ~ 0.001), but only a modest hyperpolarizing shift in the spike threshold (control −38.8 mV ± 4.4 mV versus −41.6 ± 3.8 mV in oxotremorine, P = 0.0351). Fig. 5 A–C shows the effects of oxotremorine on the FI curve of GM neurons. Like serotonin, oxotremorine increases the response to a single pulse of depolarizing current. Nonetheless, the FI curves show a marked difference in slope, with little or no change in the x intercept. Taken together, these data show that while both serotonin and oxotremorine enhance the excitability of GM neurons, they have characteristic and different effects on the FI curves of the neurons.

Open in a separate window

Fig. 5.

Oxotremorine excites isolated GM neurons and increases the number of single-spike bursters in reciprocally inhibitory networks. (A) Voltage recordings from a GM neuron in control (blue) and oxotremorine (green). Horizontal lines on traces indicate −40 mV. Current traces are shown in black; horizontal line indicates 0 nA. (B) FI plot from preparation shown in A. (C) Mean FI curve calculated as described in Fig. 2C (n = 14). (D) Distribution of bursting circuits in control (Left) and oxotremorine (Right). Grayscale corresponds to the fraction of half-center networks at each g_syn–g_h parameter set across 8 preparations. Oxotremorine (10⁻⁵ M) increased the fraction of bursting circuits at many map locations and expanded the range of g_syn and g_h values that produced half-centers. (E) Circuit frequency in 10⁻⁵ M oxotremorine as a function of circuit frequency in control saline. Asterisks indicate single-spike bursting networks as seen in Fig. 4B. Dashed line indicates x = y.

Oxotremorine produced map expansion (n = 8) and increased the percentage of bursting networks from 21.6 ± 14.6 in control to 57.2 ± 29.5 in oxotremorine (P = 0.019) (Fig. 5D). The mean burst frequency went from 0.05 ± 0.004 Hz in control to 0.34 ± 0.03 Hz in oxotremorine (P = 0.04). Fig. 5E shows an overview of the effects of oxotremorine on all of the networks. Again, considerably more points are above the line or on the line than below the line. Thus, although serotonin and oxotremorine shift the FI curves of isolated GM neurons differently, the overall conclusions taken from them are quite similar.

These dynamic clamp experiments have the precision and transparency of a modeling study, in that we were able to generate circuits in which the synaptic strength and I_h conductances are known and reproducible across preparations. The biological neurons, however, retain the variability and noise inherent in biological systems. Therefore, it is not surprising that the position and size of the half-center maps vary across experiments. Nonetheless, despite the variability of the individual neurons, and the underlying variations in circuit parameters, the effects of both serotonin and oxotremorine were consistent. Indeed, the population effects reported here in response to these substances are as robust as is usually found in the responses of neuronal preparations to neuromodulators (14–21, 36).

These data demonstrate that the responses to a neuromodulator (or by extension, other perturbations) can be statistically reproducible and reliable across circuits producing similar behavior even when their underlying parameters are quite disparate. This result is contrary to the intuitions of some (12), despite the fact that common sense tells us that it is extremely unlikely that all human brains could have identical circuit architectures that are also tuned identically. The present study also demonstrates that some specific sets of circuit parameters give rise to circuits that respond anomalously to a neuromodulator. Interestingly, there is no obvious logic to where on the parameter maps these anomalous parameter sets are found, indicating that even circuits formed from 2 neurons, with all of the noise characteristic of biological neurons, have sufficiently complex dynamics to preclude simple predictions of their behavior across all parameter regimes.

The circuits found in populations of “normal” individuals must vary in underlying structure (1). All of these circuits may work adequately and appropriately under most conditions, and most, but not all, may respond consistently to changes in neuromodulatory environment. Nonetheless, some individuals will respond anomalously to drugs (37), food, or sensory stimuli, even when their baseline behavior may be within the normal range. Untoward responses to pharmacological agents may occur when individuals contain circuits whose underlying parameters combine to make them particularly susceptible to certain kinds of perturbations, although each of the underlying parameters could be within normal ranges.

The conclusions presented here are similar to those emerging from studies of intracellular signaling networks (38, 39). Widely disparate solutions that produce similar circuit behaviors may be evolutionarily advantageous, or even necessary for evolution (38–40). Because multiple solutions are possible during evolution, genetic change could accumulate without obvious phenotypic change. The effects of these accumulated changes might be revealed only after a novel perturbation or after an additional genetic alteration occurs. A population of individuals, all of whom have circuits that are “good enough” to allow each individual to function under usual environmental conditions, but who differ among themselves in their underlying structure, is more likely to contain certain individuals who are well-suited to respond adaptively to a novel stimulus than a uniform population. Maintaining in the population individuals with anomalous responses to an environmental or internal perturbation may be necessary to ensure sufficient diversity to provide the population with the capacity to cope with a variety of perturbations, even if not all individuals respond the same way.

Materials and Methods

Crabs (C. borealis; n = 38) were purchased from Yankee Lobster and kept in artificial sea water until used. Animals were chilled for 30 min on ice, and then the STG was dissected and pinned into a Sylgard-coated (Dow Corning) dish in 10–12 °C saline. Saline consisted of 440 mM NaCl, 11 mM KCl, 13 mM CaCl₂, 26 mM MgCl₂, 11.2 mM Trizma base, and 5 mM maleic acid, pH 7.47.

Supplementary Material

Acknowledgments.

Footnotes

References